Meet TamaGROTchi: The first fully observable digital pet

Why did original digital pets fail? No telemetry, no alerting, no incident response.

Now imagine if they had:

• Metrics for hunger, mood, energy
• Logs when they evolve
• Alerts routed through IRM
• On-call rotations for caretakers

Meet TamaGrotChi: the fully observable digital pet. TamaGrotChi is the first fully instrumented digital lifeform. This WiFi-enabled pet is powered by OpenTelemetry and reports its hunger, happiness, and emotional instability straight into Grafana Cloud.

Every button press is a trace.

Every ignored cry for attention is a log.

Every missed feeding window is a measurable SLO violation.

If you think you understand alert fatigue now, wait until Grot starts escalating to IRM because you breached the Mean Time To Feed.

Chris Lawliss (00:00):

Hello GrafanaCON, I'm Chris.

Nick Downton (00:00):

And I'm Nick. And we're here to introduce you to TamaGROTchi. Now show of hands, who had a Tamagotchi growing up? Who had a Tamagotchi that died growing up? Who had a Tamagotchi that died, and you don't know why it died. [Chris Chuckles] Yeah, that seems like most people that had a Tamagotchi had the same problem. Well, TamaGROTchi is the new 2026 version, that is from us here at Grafana. Grot's our dinosaur mascot, now he can take him with you and he talks to you. We've given him a cell phone so when he gets hungry, he can even slack you. Grafana runs quarterly hackathons. We won this last hackathon for the science fair. Most past winners have maybe added something to the product. You've maybe heard of AI observability, or the Grafana assistant. Well, we decided to make a cute little pet. TamaGROTch is built with an ESP 32 board and transmits logs, traces and metrics via hotel.

(01:11):

Whether you're feeding or playing with them, each interaction is sent as one of those.

Chris Lawliss (01:18):

Cool. So it's all fun and games, but actually a lot of our customers, they're not actually using Kubernetes, they can't use a Java SDK, they're actually building observability onto small devices like this. So we have customers who have robots running around warehouses. They're building EPOS terminals. They're fairly hardware restrained environments with limited memory. They might not have great internet connections and they can't afford to be running things like the hotel collector or alloy, and a full SDK on those devices. So we kind of wanted to experiment and see how easy is it just to send raw Otel data. So we're just creating a JS on packet, sending that over HTTP and sending that straight to Grafana Cloud. Yeah. So the other thing that we wanted to kind of do is our moms don't know what we do at Grafana. So I'm an Observability Architect, I talk about observability all day.

Nick Downton (02:25):

I'm a data engineer, I deal with data loading scripts.

Chris Lawliss (02:29):

Yeah, but this was the only way we could actually tell our families what we do at Grafana. So when we were building these, one of the things that we needed to decide was what telemetry signals we were gonna use. So the trap is just send everything as a log or just send everything as a metric, whatever you are used to. So we went down that route and we had our drop sick metric, and that was a zero or one, and we could quite easily build a dashboard that counted how many sick TamaGROTchis we had, but we weren't then able to answer other questions like, how did they get sick? Who let them get sick? So we actually started implementing things like logs. So every button press that is a log. We've got tracing as well, and that let us build out a much better suite of dashboards.

(03:24):

So how do we get this data? How do we observe it? So who here knows knowledge graph or entity graph? Perfect. So yeah, we have written a custom entity for Knowledge Graph. Usually this would be a top down view of your, full application stack. So you'd see your Kubernetes pods, your nodes, your application workers, et cetera. We just have TamaGROTchis, but we have SLOs that are custom to our use case. So you can see things like when they're hungry, when they're ill. Those are all custom built into knowledge graph. Other things we've got, so we've got tried and tested, sort of SRE conventions in place. So SLOs, we have meantime to feed. So we're, we're looking at how many times TamaGROTchi is requesting to be fed and how long it takes for them to actually get responded to, and then we've got our red metrics as well, which are obviously really important.

(04:29):

So to dive deeper into the data. So Knowledge Graph gave us that high level overview. We've also got a bunch of custom dashboards that allow us to see the individual metrics, the logs that are coming through, and then we can dive into a single pets dashboard, and I think no talk at GrafanaCON would be complete without a bit of AI in there. So we've also given the Grafana assistance, some capabilities, we've made some custom, custom scenes in there where it understands what a TamaGROTchi is. It knows that we need to keep it alive, and actually he gets pretty annoyed when you neglect him.

Nick Downton (05:15):

Yeah, and it really gives a new name to postmortem. [Chris Laughs] Yeah, so come by and see us at our booth at the science fair. you can stop by the IoT booth as well for a chance to take one home via the NFC raffle. We want to show you how the dashboards work. It's a great learning tool, as Nick has already said. So how I taught my mom what logs, traces, and metrics are. [Nick Chuckles] Yeah, and come by and play with 'em. They're sure are a lot of fun.

Chris Lawliss (05:47):

Yeah, any questions? We're down at the science fair, I hope to see you guys there. Cool, thank you. [Crowd Cheers]

Nick Downton (05:52):

Thank you.